基于SVM整定PID控制参数的网络拥塞控制AQM算法  

NETWORK CONGESTION CONTROL AQM ALGORITHM BASED ON SVM TUNED PID CONTROL

作　　者：胡鹏莎[1] 张浩平[1] 丁文[1] 徐波[2] 

机构地区：[1]南京林业大学信息科学与技术学院,江苏南京210037 [2]南京林业大学信息网络中心,江苏南京210037

出　　处：《计算机应用与软件》2012年第8期101-105,共5页Computer Applications and Software

基　　金：教育科研重大应用示范项目(8509000030);南京林业大学创新基金项目(X2011-11)

摘　　要：针对PID主动队列管理(AQM)中参数自整定、无法适应复杂的非线性网络环境等缺点,提出一种基于最小二乘支持向量机(LS-SVM)与PID控制相结合的智能AQM算法——LSPID算法。仿真实验表明,LSPID算法具有更好的收敛性,能将队列长度更加快速地收敛到期望值附件,在动态拓扑、链路容量变化、有突发流的网络环境中能很好地保持稳定性和鲁棒性,其控制效果比专家智能PID和神经网络PID主动队列管理算法都更优越。Focus on the disadvantages of PID active queue management （AQM） in self-tuning, cannot adapt to complex nonlinear network environment, etc. , this paper puts forward an intelligence AQM algorithm--LSPID, which is based on the combination of least squares support vector machine （LS-SVM） and PID control. Simulation experiments show that the LSPID algorithm has better convergence, it can more quickly converge the queue length to approximate expectations, can maintain very good stability and robustness in network conditions with dynamic topology, link capacity change and sudden flow, and the control effect is significantly better than two other AQM algorithms, including the expert intelligence PID and the neural network PID.

关 键 词：主动队列管理 专家智能 神经网络 最小二乘支持向量机 PID 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]